Ecological validity of a deep learning algorithm to detect gait events from real-life walking bouts in mobility-limiting diseases

Robbin Romijnders*, Francesca Salis, Clint Hansen, Arne Küderle, Anisoara Paraschiv-Ionescu, Andrea Cereatti, Lisa Alcock, Kamiar Aminian, Clemens Becker, Stefano Bertuletti, Tecla Bonci, Philip Brown, Ellen Buckley, Alma Cantu, Anne Elie Carsin, Marco Caruso, Brian Caulfield, Lorenzo Chiari, Ilaria D'Ascanio, Silvia Del DinBjörn Eskofier, Sara Johansson Fernstad, Marceli Stanislaw Fröhlich, Judith Garcia Aymerich, Eran Gazit, Jeffrey M. Hausdorff, Hugo Hiden, Emily Hume, Alison Keogh, Cameron Kirk, Felix Kluge, Sarah Koch, Claudia Mazzà, Dimitrios Megaritis, Encarna Micó-Amigo, Arne Müller, Luca Palmerini, Lynn Rochester, Lars Schwickert, Kirsty Scott, Basil Sharrack, David Singleton, Abolfazl Soltani, Martin Ullrich, Beatrix Vereijken, Ioannis Vogiatzis, Alison Yarnall, Gerhard Schmidt, Walter Maetzler

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review


Introduction: The clinical assessment of mobility, and walking specifically, is still mainly based on functional tests that lack ecological validity. Thanks to inertial measurement units (IMUs), gait analysis is shifting to unsupervised monitoring in naturalistic and unconstrained settings. However, the extraction of clinically relevant gait parameters from IMU data often depends on heuristics-based algorithms that rely on empirically determined thresholds. These were mainly validated on small cohorts in supervised settings. Methods: Here, a deep learning (DL) algorithm was developed and validated for gait event detection in a heterogeneous population of different mobility-limiting disease cohorts and a cohort of healthy adults. Participants wore pressure insoles and IMUs on both feet for 2.5 h in their habitual environment. The raw accelerometer and gyroscope data from both feet were used as input to a deep convolutional neural network, while reference timings for gait events were based on the combined IMU and pressure insoles data. Results and discussion: The results showed a high-detection performance for initial contacts (ICs) (recall: 98%, precision: 96%) and final contacts (FCs) (recall: 99%, precision: 94%) and a maximum median time error of −0.02 s for ICs and 0.03 s for FCs. Subsequently derived temporal gait parameters were in good agreement with a pressure insoles-based reference with a maximum mean difference of 0.07, −0.07, and <0.01 s for stance, swing, and stride time, respectively. Thus, the DL algorithm is considered successful in detecting gait events in ecologically valid environments across different mobility-limiting diseases.

Original languageEnglish
Article number1247532
JournalFrontiers in Neurology
StatePublished - 2023


FundersFunder number
Land Schleswig-Holstein
Tyne Hospitals NHS Foundation Trust
Wellcome Trust Clinical Research Facility
Canine Research Foundation
European Federation of Pharmaceutical Industries and Associations
National Institute for Health and Care Research
Newcastle University
Deutsche ForschungsgemeinschaftB9
Newcastle upon Tyne Hospitals NHS Foundation Trust
Horizon 2020
Innovative Medicines Initiative820820
NIHR Newcastle Biomedical Research Centre
NIHR Imperial Biomedical Research Centre


    • deep learning (artificial intelligence)
    • free-living
    • gait analysis
    • gait events detection
    • inertial measurement unit (IMU)
    • mobility


    Dive into the research topics of 'Ecological validity of a deep learning algorithm to detect gait events from real-life walking bouts in mobility-limiting diseases'. Together they form a unique fingerprint.

    Cite this